The semantics and pragmatics of misrepresentation in English - a relevance-theoretic approach

(1)

The Semantics and Pragmatics of Metarepresentation in English:

A Relevance-Theoretic Approach

Eun-Ju Noh

Thesis submitted in partial fulfillment

of the requirements for the degree of Ph.D.

University College London

1998

(2)

For my parents

(3)

Abstract

This thesis deals with the nature of metarepresentation in language. It proposes linguistic-semantic and pragmatic analyses of a variety of metarepresentational expressions in English, using the framework of relevance theory (Sperber and Wilson

1986/1995). The main aim is to deepen the relevance-theoretic analysis of

metarepresentation, to apply it to a range of data which have not been previously analysed in this framework, and to compare the resulting account with alternative semantic and pragmatic accounts.

Chapter 1 looks at various types of quotation (including mention, reported speech and thought, and mixed quotation) and surveys some of the problems encountered by traditional and more recent alternative accounts; the chief problem being that they either do not acknowledge the range and variety of semantic indeterminacies in quotation, or do not provide an adequate account of how these indeterminacies are resolved during utterance comprehension. Chapter 2 introduces relevance theory and shows how the comprehension strategy it provides can be used to resolve the various indeterminacies in quotation. It also shows how the relevance-theoretic notion of metarepresentation (representation by resemblance) can be applied not only to paradigmatic cases of direct and indirect quotation, but also to a range of other cases involving the exploitation of linguistic or conceptual resemblances. What is common to all these cases is that a representation is used with a guarantee of faithfulness to some other representation, rather than truthfulness to some state of affairs. The claim that a metarepresentation can be faithful enough without being identical to the original is illustrated and explored.

(4)

(5)

Acknowledgements

I owe a debt of profound gratitude to numerous individuals and institutions for their assistance in the writing of this thesis.

First, I am deeply grateful to my supervisor Professor Deirdre Wilson, who has been a constant source of inspiration and stimulation throughout the preparation of this thesis. She has commented on a number of earlier versions of each chapter, which she always returned full of invaluable advice and constructive criticism. Thanks to her help, I have participated in many conferences and published papers in journals and a book, all of which has been a good foundation for this thesis. She has helped me not only with my study but also with my life in London. I will never forget the moment when she pointed at Korea on a world map and showed her interest in the place where my family lived, in order to cheer me up when I was missing them. She has been a helpful supervisor, a reliable guide, a skilful counsellor and a trustworthy friend.

I am also very grateful to my second supervisor Dr. Robyn Carston. I acknowledge that my interest in metarepresentation arose purely from the discussions I had with her when she was my MA thesis supervisor. She read earlier versions of the thesis and has guided me with insightful comments on various aspects. She also gave me the chance to collaborate on a paper with her and participate in a conference together. Her help was never limited to the academic field; without her warm friendship, I could not have survived the difficult time in London after my family went back to Korea.

(6)

My thanks also go to the Department of English Teaching and the Department of English Language and Literature of Seoul National University, where I studied before I came to UCL; particularly to Professor Choon-Hak Cho, Professor Nam-Sheik Park, Professor Ki-Sun Hong and Professor Chung-Min Lee for their teaching and caring advice.

I am also grateful to the British Council in Seoul, the Committee of University Vice Chancellors in the United Kingdom, and the Graduate School of UCL for granting various funds, which have been of great help to my research at UCL.

I am very grateful to my parents, who have always supported me and particularly taken care of Chi-Young and Eun-Young whenever I could not afford to. It is not just because of their love, but because of their great sacrifice that I was able to finish this thesis. I also thank my sisters and brother, Mi, Nam, Kyong-Hee, Eun-Soo, Eun-Soon for their love and care. I also want to express my gratitude to my parents-in law, and relatives and friends, especially Sook Lim and Sun-Jung Kim, who have been always willing to help me and encourage me.

(7)

Semantic contradiction and pragmatic reanalysis 157 3.2. 1.3 Problems with standard accounts 161 3.2.2 Van der Sandt (1994): Denials and the echo operator 163 3.3 A Previous Relevance-Theoretic Account of Metalinguistic Negation 173 3.3.1 Carston (1994/1996): Implicit echoic use 173 3.3.2 Metalinguistic negation without echoic use 178

3.4 Metarepresentational Use in Negation 180

3.4. 1 Varieties of metarepresentational use in negation 180 3.4.2 The grounds for using metalinguistic negation 185 3.4.3 Truth-conditional content of metal inguistic negation 189

3.5 Summary and Conclusion 195

Appendix to Chapter 3: Metalinguistic Negation in Korean 197

Chapter 4: ECHO QUESTIONS

4.2 Standard Accounts of Echo Questions 208

4.3 Do Echo Questions Require a Prior Utterance? 210 4.3.1 Echo questions without a prior utterance 211

4.3.2 Echo questions reformulated 214

4.4 Are Echo Questions Cases of Metalinguistic Use? 217 4.4.1 Echo questions and metalinguistic negation 217

4.4.2 Horn (1989) and Cooper (1983) 220

4.4.3 Metarepresentational use m echo questions 223

4.5 A Relevance-Theoretic Account of Echo Questions 225

4.5.1 Blakemore (1994) 225

4.5.2 Echo questions as attributive metarepresentations 230

4.5.3 Echo questions as interrogatives 238

(10)

Chapter 5: METAREPRESENTATIONAL USES IN CONDITIONALS

5. 1 Introduction 242

5.2 A Survey of Accounts of Conditionals 243

5.3 Metarepresentational Use of Antecedents 249

5.3.1 'Given' antecedents in the literature 250

5.3.1.1 'Givenness' in general 250

5.3.1.2 Accounts of 'given' antecedents 251 5.3.1.3 'Given' antecedents and speaker's knowledge 253

5.3.2 Metarepresentational antecedents 259

5.3.2.1 Varieties of metarepresentational antecedents 259

5.3.2.2 Truth conditions of metarepresentational antecedents 262

5.4 Metarepresentational Use of Consequents 264

5.4.1 Introduction 264

5.4.2 Speech-act accounts 265

5.4.2.1 Speech-act conditionals 265

5.4.2.2 Metalinguistic conditionals 269

5.4.2.3 Some problems with speech-act accounts 272

5.4.3 A metarepresentational account 274

5.4.3.1 Metarepresentational consequents 274

5.4.3.2 Truth conditions of metarepresentational consequents 279

5.5 Metarepresentational Use in Antecedent and Consequent 283

5.5.1 Metarepresentational uses in both clauses of a conditional 283 5.5.2 Counterfactual indicative conditionals 284 5.5.3 Horn's metalinguistic metaphorical conditionals 286

5.6 Summary and conclusion 287

CONCLUSION 289

(11)

Introduction

Metarepresentation in Communication

Humans have the ability to construct mental and public representations of many different types. Some are sensory: we can represent in our mind the sound of church bells, the appearance of the beautiful roses we saw this morning, and the taste of the food we had in Italy. We can publicly represent them in a recording, in a painting, in a photo, etc. Others are conceptual, logical or mathematical: we can represent in our mind a thought, a logical argument or a mathematical proof. We can publicly represent them in an utterance, a dissertation or a mathematical formalism. According to Fodor (1983), the mind must have several different systems of representation and computation which together account for these abilities.

Humans also have the ability to construct mental and public representations of these representations. We can represent the utterances made by our friends, the news we listened to on TV, or the ideas of some philosopher we have read. This metarepresentational ability has received much attention in recent work in philosophy and psychology, where the focus has been on the so-called 'theory-of-mind', i.e. the ability to explain and predict the behaviour of others by attributing to them certain beliefs, intentions and desires. (See the contributions in Astington, Harris and Olson (eds.) 1988; Byrne and Whiten (eds.) 1988; Carruthers and Smith (eds.) 1996; Mind

and Language 13.1 (1998); Carruthers and Boucher (eds.) 1998; and Wellman 1990;

Gopnik 1993; Leslie 1994).

(12)

Introduction

Baron-Cohen et al. 1985; Happé 1992). Despite various disagreements among existing accounts about how these abilities are best explained, it is generally agreed that young children who have not reached the stage of having the theory-of-mind and autistic people lack the theory-of-mind ability. It is also generally agreed that full-fledged inferential communication in humans draws on this same ability to attribute underlying beliefs, intentions and desires to others (Grice 1989; Sperber and Wilson

1986/1995; Sperber 1996). This is one way in which metarepresentational abilities

play a role in human communication. I will look at it briefly before moving on to a second type of metarepresentational ability, which will be my main concern in this thesis.

Grice's (1989) inferential account of utterance interpretation is based on the hearer's ability to metarepresent a communicator's intention. In Grice's framework, the hearer does this by assuming that the speaker intended her utterance to satisfy a Co-operative Principle and maxims of truthfulness, informativeness, relevance and clarity. Consider (1):

(1) Peter: Where are you going? Mary: We need coffee.

According to Grice, in (1), Mary's reply is not as informative as is required. Assuming that Mary is obeying the Co-operative Principle and maxims, Peter looks for an interpretation that satisfies them: for example, he assumes that she is going to buy some coffee; and he attributes to Mary the intention to convey this interpretation. Thus, Grice's Co-operative Principle and maxims might be seen as an integral part of the 'theory-of-mind' ability as applied to the production and interpretation of utterances.

However, Grice's account of inferential communication does not fit straightforwardly into a metarepresentational 'theory-of-mind'. In the first place, 'theory-of-mind' abilities are generally seen as innately determined, unconscious and modularised, whereas Gricean inferential comprehension looks more like a conscious reasoning process and the source of the Co-operative Principle and maxims is

(13)

debatable. It has also been argued that the Co-operative Principle and maxims are descriptively and explanatorily inadequate, and should be replaced by a communicative principle that fits better with general assumptions about human cognition. (Sperber and Wilson 1986/1995; Sperber and Wilson 1987b).

In the second place, it is not clear how Grice's account of inferential communication fits with what is known about the development and breakdown of 'theory-of-mind' abilities (see references above). It is well established that these abilities develop over time, and that children and autistic people in whom they are not fuliy developed can engage in some degree of inferential communication; but it is not clear how Gricean inferential communication can be achieved to a greater or lesser degree. Sperber and Wilson (1986/1995: chapter 1, sections 11 - 12) attempt to deal with these issues by attributing to communicators two types of intention, involving different degrees of metarepresentational ability: an informative intention, and a higher-order communicative intention (i.e. an intention that the hearer should recognise the informative intention). They argue that some degree of communication may be achieved as long as the informative intention is recognised, even if the ability to metarepresent the higher-order communicative intention is not there.

Sperber (1994a) develops these ideas, arguing that even the ability to recognise the basic informative intention may involve different degrees of metarepresentational ability in different circumstances. I will illustrate this claim using Grice's maxims of truthfulness, 'Try to make your contribution one that is true; do not say what you believe to be false,' as applied to example (1) above (where the informative intention is the intention to inform Peter that they need coffee and that Mary is going to get some).

Suppose that, in (1), Peter has bought some coffee on his way home, so that there is no need to buy coffee. If Peter expects Mary always to say something that is actually true, he will have to reject the most straightforward interpretation of Mary's utterance, as intended to inform him that they need to buy coffee. This is not, of course, what happens, and Grice's maxims of truthfulness allow for this. Using these maxims, Peter may interpret Mary's utterance correctly, not by looking for what actually is true, but by looking for what she might have thought would be true. As

(14)

introduction

Mary did not know that Peter had bought some coffee, she might have thought that they needed coffee and intended him to believe that she was going to buy some coffee. In order to arrive at this intended meaning, he needs to attribute to Mary the false belief that they need coffee. This is a theory-of-mind ability which is lacking in children below a certain age, who will therefore be unable to use Grice's maxims of truthfulness, and should make characteristic mistakes in comprehension. In this way, development of communicative abilities is clearly linked to 'theory-of-mind'.

Consider now the case of lying. Here, the speaker covertly violates the maxim of truthfulness by saying something she believes to be false. Grice handles this by saying the speaker covertly violates the maxim of truthfulness. What he does not explain is how the hearer can sometimes understand the intended interpretation, even when he knows the speaker is lying. Consider (1) again. Suppose Mary knows that Peter bought some coffee, but she does not know that Peter knows that she knows it. Then Mary's utterance in (1) is a lie. If Peter knows this too and assumes that she always says something she believes to be true, he will be unable to identify the intended meaning. If he does not have this expectation, he may be able to arrive at the intended interpretation using a more complex metarepresentation of the form: 'Mary intends me to believe that she believes P (= We need coffee)'. This strategy involves a complex, multi-layered metarepresentational inference, which will again be beyond the capabilities of people at certain developmental stages in 'theory-of-mind'.'

In this thesis, I will be concerned with a different type of metarepresentational ability, which has been much less studied in the 'theory-of-mind' literature (though see Happd 1993). This appears in cases where the communicated information (the 'message') itself contains a metarepresentational element, which is intended to be recognised as such. Consider (2):

(15)

'I

(2) a. Mary believes that I sent the letter. b. Mary said that I had sent the letter. c. It is false that I sent the letter.

Here, three types of representation with conceptual content are metarepresented: a belief in (a), an utterance in (b) and a proposition in (c). A belief is a mental representation of the type primarily focused on in the literature on theory-of-mind. An utterance is a public representation, and a proposition is an abstract representation. As these examples show, all three types of metarepresentation can appear in the communicated content of an utterance. The main aim of this thesis is to provide an account of how such metarepresentational utterances are understood.

The examples in (2) metarepresent the content of the original representation, but it is also possible to metarepresent various aspects of form. Consider (3):

(3) a. Mary said 'She sent the letter'.

b. Mary said 'He seems like to think like it's like a bit like squishy'.

By directly quoting Mary's utterance, the speaker of (3) gives the hearer some idea of the form of the original. The purpose may be merely to inform the hearer of what Mary said, or it may be to express an attitude, e.g. to Mary's frequent use of 'like' in (3b).

In (2) and (3), the presence of a metarepresentational element is overtly indicated by use of verbs such as 'believe' or 'say'. However, metarepresentation is not always overtly indicated. Consider (4):

(4) Jane: What did Mary say?

Peter: He seems like to think like it's like a bit like squishy.

(16)

Introduction

I also want to provide an account of how the presence of a metarepresentational element may be inferred.

In relevance theory (Sperber and Wilson

1986/1995), metarepresentation is

defined as the use of one representation to represent another in virtue of some resemblance between them, whether in content or form. My main aim in this thesis is to apply the relevance-theoretic notion of metarepresentation to various types or uses of linguistic metarepresentation, and provide an explanation of how they work.

Chapter 1 looks at various types and uses of quotation: direct and indirect quotation, pure quotation (mention), and mixed quotation. These vary as to the type and source of the original, the degree of resemblance between original and quotation, the purpose of using the quotation, etc. not all of which are linguistically marked. The chapter surveys both traditional and more recent accounts of quotation, and argues that they do not pay enough attention to indeterminacies in interpretation, or provide a pragmatic framework in which these indeterminacies can be resolved.

Chapter 2 starts with a brief outline of relevance theory and the relevance-theoretic comprehension strategy. It introduces the relevance-relevance-theoretic notion of metarepresentation and shows how some paradigmatic cases of quotation would be dealt with on a relevance-theoretic account. The account is then applied to a range of further data not generally considered in the literature on quotation, which are also cases of metarepresentational use. The truth conditions of metarepresentational utterances are briefly considered, and analysed in terms of the notion of pragmatic enrichment (Sperber and Wilson 1986/1995; Carston 1998a). In the Appendix, the account is applied to the Korean dialect expression 'kesiki'.

(17)

Introduction

position of Carston (1994/1996; 1998dm press). In the Appendix, Korean negation is briefly investigated from the point of view developed in the chapter.

Chapter 4 explores echo questions. It looks at previous treatments of echo questions, both inside and outside relevance theory, and extends up the relevance-theoretic analysis to deal with some standard and non-standard types of echoic question. Echo questions are analysed as ordinary interrogatives involving metarepresentational use, whose propositional form is the result of pragmatic enrichment. The account is briefly compared with the position of Blakemore (1994).

(18)

Chapter 1 Analyses of Quotation

1.1 Introduction

Linguistic metarepresentation, as I have defined it, involves the use of an utterance (a public linguistic representation) to represent some other representation than a thought of the speaker's at the current time. In the existing literature, two types of linguistic metarepresentation have been extensively discussed: mention (or 'pure quotation'), in which a linguistic expression is used to represent the abstract type of which it is a token, and reported speech and thought, in which an utterance is used to represent an utterance or thought attributed to someone other than the speaker at the current time. This chapter will look at these two central cases of quotation, and survey some existing semantic and pragmatic analyses, as a background to the relevance-theoretic account which I develop in chapter 2.

Some typical cases of pure quotation or mention are given in (1):

(1) a. 'Life' is monosyllabic. b. 'Life' has four letters. c. 'Life' is a noun.

In (1), 'life' is not used to convey one of its standard dictionary meanings. Rather, it refers to itself, or, more precisely, to the word type of which it is a token. 'Life' in (a) refers to the word type as a spoken form, in (b), as a written form, and in (c), as an expression belonging to a certain grammatical category. An adequate account of mention should describe and explain how such uses of language work.

(19)

(2) a. Socrates said,

'Know

yourself!'

b. Socrates said that we should know ourselves.

(2a) is a case of direct speech, which preserves many of the formal as well as the semantic properties of the original utterance being quoted. (2b) is a case of indirect speech, which preserves the semantic properties, but not necessarily the formal properties, of the original utterance. Another commonly discussed type of quotation is free indirect speech, which has similarities to both direct and indirect speech, but also differs from both. These three types of quotation differ from mention (pure quotation) in that they are used to represent attributed utterances (or thoughts), rather than abstract linguistic types. An adequate account of quotation should describe and explain how such examples work.

My first aim in this chapter is to survey the various types and uses of quotation. Some of these have been widely discussed in the literature, but others have not. At the same time, I will look briefly at some existing analyses of quotation, drawing attention to some features which I believe should be shared by any adequate account, and to others which I believe should be replaced. I group these accounts into two categories: the traditional and the recent. Recent accounts remedy many of the weaknesses of more traditional accounts, but I will argue that they still leave some work to be done.

(20)

I

Recent accounts of quotation avoid these weaknesses. They deal with mention and reported quotations together, and argue that some types of quotation may not have the same semantic or formal properties as the original. My main concern in this chapter is with these recent approaches. I will argue, first, that they need to recognise that resemblance rather than identity is the very essence of quotation, and second, that for lack of an adequate pragmatic framework, they are more descriptive than explanatory. In chapter 2, I will develop an account designed to remedy these defects.

This chapter is organised as follows: In section 1.2, I will survey various types and uses of quotation. In section 1.3, I will look at some traditional accounts of the use-mention distinction, and of reported speech and thought. In 1.4, I will look more closely at recent alternative approaches to quotation, pointing out some of their strengths and weaknesses. Section 1.5 is a summary and conclusion.

1.2 Varieties of Quotation

(21)

1.2.1 Pure quotation (Mention)

One reason for mentioning a linguistic expression is to comment on it. Consider (3), repeated from (1):

(3) a. 'Life' is monosyllabic. b. 'Life' has four letters. c. 'Life' is a noun.

As we have seen, in (3), 'life' is used to refer to the abstract linguistic type of which it is a token, specifically to the spoken form in (a), to the orthographic form in (b), and to the grammatical category in (c). In each case, part of an utterance is being used to represent an abstract representation, which may be seen as having purely linguistic properties (assuming that spelling is a purely linguistic property). These are the classic cases of mention, discussed in all accounts.

However, what is mentioned need not be an abstract linguistic representation, and the reason for the mention may go beyond the purely commentative. Consider (4), written on the Kellogg's box, and (5), on an instant national lottery ticket (in the U.K.):

(4) If you don't see

Kellogg's

on the box, it isn't

Kellogg's

in the box.

(5) MATCH 3 TO WIN OR MATCH 2 AND 'DOUBLE' TO DOUBLE YOUR WIN

OR MATCH 2 AND 'TRIPLE' TO TRIPLE YOUR WIN

(22)

than a token. These contrast with the other uses of

'Kellogg's' and 'DOUBLE' or

'TRIPLE', to which the shapes are immaterial.

Not only words, but also phrases or sentences can be mentioned. Consider (6) -(8):

(6)

'Kick the bucket' has three words.

(7) 'Regent's Park is near UCL' is a sentence of English.

(8) 'It's spring. Weather couldn't be better. I have stayed in my room for three days' has a different subject in each sentence.

In (6) - (8), a phrase, a sentence, and a group of sentences are mentioned, respectively. In the same vein, a single letter can be mentioned, as in (9):

(9) 'G' is a consonantl a capital letter! difficult to pronounce! the seventh letter in alphabet! my favourite letter! stands for the fifth musical note

In each case, what is metarepresented is an abstract object with linguistic properties, which may be singled out and drawn to the audience's attention.

In the literature, mention is generally seen as a case where a linguistic expression is used to refer to itself. I have already suggested that this needs some qualification. It is also worth pointing out that the properties of the original that are singled out may be rather far removed from those of the quoting expression. For example, we can represent a word's pronunciation by using its written form, as in (3 a), and its written form by using its spoken form, as in the spoken version of (3b).

It is also generally assumed that mention always involves the representation of formal properties, as in the above examples. However, there is no obvious reason why content cannot be mentioned. Consider (10) - (11):

(23)

(11) 'Life' means 'a manner of life' in this context.

In (10), what is picked out by mentioning the two sentences is not their form but their content, including semantic and logical properties. In the same way, in (11), it is the meaning of 'manner of life' that is mentioned, rather than its form. These cases are very similar to standard cases of mention, in that their originals are abstract linguistic representations. I propose to include them under mention, as cases of mentioning a meaning.

There are some more obvious cases of mentioning a meaning, which involve explicit indication that a proposition is involved, as in (12):

(12) On this occasion 'I am here' expressed the proposition 'Eun-Ju Noh is in London on the 14th of June, 1998'.

Utterances may also be used to represent mathematical formulae and logical arguments. Further examples involving mention of non-linguistic abstract representations are given in (13) - (14):

(13) '13' is a sinister number in this culture.

(14) '*' marks ungrammaticality in syntax.

The number '13' in (13) and the sign '' in (14) should also be treated as cases of mention. So the types of abstract representation that can be mentioned go well beyond the purely linguistic.

(24)

r

(15) A: That's a rhinoceros. B 1: Spell it out for me. B2: Don't give it a biscuit.

According to Levinson, 'it' in (Bi) is a discourse-deictic pronoun, which refers to a linguistic expression, whereas 'it' in (B2) is a regular anaphoric pronoun, which refers to the same entity as a prior linguistic expression. 'It follows that there is a close, but quite unexplored, relation between discourse deixis and mention or quotation' (Levinson 1983: 86).

Davidson (1979) makes a similar point, with examples that might be seen as midway between regular mention and discourse deixis. Consider (16), from Davidson (1984: 81) (my own emphasis), and (17):

(16) Dhaulagiri is adjacent to Anapurna, the mountain whose conquest Maurice Herzog described in his book of the same name.

(17) Barbarossa was so-called because of his red beard.

Davidson analyses these examples as mixed cases of use and mention. The idea is that to give the truth-conditions of (17), for example, one must produce a paraphrase containing one use and one mention of the same expression, e.g. 'Barbarossa was called "Barbarossa" because of his red beard'. These examples, like Levinson's, are interesting, because they suggest that ordinary use of a linguistic expression makes the abstract type easily available for later exploitation at either the referential or the truth-conditional level. In metarepresentational use proper, when a single occurrence of an expression is used to represent the abstract type of which it is a token, the same easy availability is exploited, though in a slightly different way.

(25)

Chiprer I

of the examples used so far would be comprehensible without overt quote marks, especially in spoken form. Or consider another example:

(18) There is no parody, only paradise.

Suppose the speaker of (18) is looking for the entry 'parody' in the subject index of a book. In that case, 'parody' would be mentioned rather than used. But the spoken form of (18) would be straightforwardly understood without overt quote marks. An adequate account of quotation should explain how the presence of a quotation, when not overtly encoded, can be inferred.

In this section, I have looked at various types and uses of pure quotation (or mention). In pure quotation, the original is an abstract linguistic or non-linguistic representation, and what is singled out may be its form or content, its linguistic or non-linguistic properties. One question raised by this discussion is how the audience recognises the presence of a mention, the type of the original and the particular properties that are being represented. These points will be discussed further below.

1.2.2 Reported speech and thought

Another type of quotation is reported speech and thought, where what is quoted is not an abstract representation but an attributed utterance or thought. Traditionally, three types of reported speech have been distinguished: direct speech, indirect speech and free indirect speech. Direct speech is often analysed as a type of quotation involving representation of the surface form of the original utterance. Consider (19):

(26)

Chapter 1

In (19), Mary reports Peter's utterance in the form in which it was uttered: The imperative sentence type is preserved, the tense of verb, the deictic 'here' and personal pronoun 'I' are all as they were in the original.

Indirect speech is often analysed as a variety of quotation involving reproduction of the content of the original rather than its surface form. Peter's utterance in (19) might be reported in indirect speech as in (20):

(20) Mary: Peter told John to leave there at once, and never come back.

In (20), Mary is reporting the content of Peter's utterance. The report is embedded under a verb of saying, and person and deictics are all shifted from those of the original: from unexpressed subject 'you' to 'John' and from 'here' to 'there'. What form the result takes depends on context. For example, if Mary is talking to John (when he has not heard the utterance properly), she may say, 'Peter told you to leave here at once, ...'. Tense is also back-shifted, as in (21):

(21) a. A: I am going to meet my boss soon.

b. B: A said that he was going to meet his boss soon.

In (21), B's indirect speech shows that present tense 'am' is backshifted to past tense 'was'. Personal pronouns 'I' and 'my' are also shifted to 'he' and 'his'.

A central issue in the utterance on indirect speech is how to deal with the

ambiguity between de dicto and de re readings.' A classical example is that of Oedipus, who unwittingly married his mother Joca.sta:

(22) Oedipus said that his mother was beautiful.

(22) has at least two readings: On the de dicto one, Oedipus said 'my mother is beautiful'; on the de re one, he may have said 'Jocastalmy wife is beautiful' without

For representative discussion, see e.g. Quine 1956, 1960; Davidson 1968; Partee 1973; Coulmas 1986; Bertolet 1990. The literature is mainly philosophical, and the issue is not discussed in the standard linguistic and literary works on indirect speech.

(27)

unaprer I

knowing that the woman he referred to is his mother. Such examples raise many problems. The one that interests me here is the pragmatic one: how does the hearer decide what interpretation the speaker intended to convey? In connection with this, let us look at Donnellan's (1966) example:

(23) The man drinking the martini is a famous philosopher.

In (23), 'the man drinking the martini' can be used in two ways: referential and attributive. That is, it can be used to refer to a specific person under a particular description which may or may not be true, or to pick out whoever it is that fits the description, whether or not the speaker is aware of this. If someone says (23), intending to refer to a specific person who is in fact drinking water, the resulting statement may be true, despite the fact that it contains a false description. The question that interests me is how such an utterance would be reported. There are two

possible ways, given in (24):

(24) a. He said that the man drinking the martini is a famous philosopher. b. He said that the man drinking water is a famous philosopher.

I believe that in different circumstances, either of these would be possible. Or consider (25), which displays a different but parallel ambiguity involving an

expressive element:

(25) He told me that he submitted the rubbish to a journal.

In (25), the expression 'the rubbish' may be attributed to the original speaker, but is more likely used by the reporter to express an opinion of the work submitted.

(28)

Chapter 1

is directly quoting the original utterance (this is a case of mixed quotation; see next section). By the same token, (22) can be used whether or not Oedipus knew that his wife was his mother. What the hearer has to infer is whether the expression 'mother' is attributed to Oedipus or used by the reporter himself. In section 1.3.3.2, 1 will look in more detail at semantic accounts of indirect speech, where the complement clause of indirect speech has been analysed in various ways.

So far, I have looked only at cases where the reported speech is embedded under a verb of saying. This overtly indicates the presence of a quotation, and the source of the original utterance (real or imagined). There is another type of reported speech in which no such indication is given. This is the case of free direct or indirect speech. Consider (26):

(26) He came toward me in the dark. 'Who are you? Stay away.' But he kept on towards me. 'Help!'

Nobody seemed to hear me.

Passages such as those in (26) are often found in novels. (26) contains two direct speech reports which are not overtly marked as such. These are cases of free direct speech.

An example of free indirect speech is given in (27):

(27) A: What did she say?

B: We should look after ourselves.

(29)

introductory clause. In free direct and indirect speech, the presence of the quotation has to be inferred, i.e. recognised by pragmatic means.

Free direct speech raises few theoretical problems, and everyone would agree that 'free' means 'not overtly marked', e.g. by embedding under a verb of saying. However, in the case of free indirect speech, there has been more room for disagreement, because the phenomenon is much richer. Most people would agree that omission of an explicit verb of saying converts indirect speech into free indirect speech, as in (27b). Other characteristics of free indirect speech may be present even when an explicit verb of saying is used. (See also section 1.3.3.3.) Thus, consider (28):

(28) a. She asked, did he have some money? b. Did he have some money, she asked.

In (28a - b), the quotation, as in regular cases of free indirect speech, preserves the sentence type of the original, with tense and person shifted. It thus seems preferable to analyse it as a case of free indirect speech, treating the verb of saying as parenthetical.

While regular indirect quotations convert originals of interrogative form to embedded declarative clauses, free indirect speech preserves the sentence type of the original, with person and tense shifted as in (29b), where the original utterance might be (29a):

(29) a. A: Do you have some money? b. B: Had he some money?

(30)

C

(30) To-morrow was Monday, Monday, the beginning of another school week! (Lawrence, Women in Love, p.1 85)

In (30), the fact that 'tomorrow' is used to refer to a past time, and co-occurs with 'was', serves as an indicator of (free) indirect speech.

So far, I have taken for granted the assumption found in most of the literature, that the main function of direct and indirect speech is to report attributed utterances and thoughts. Notice, now, that not all repeated utterances are used to report. Many cases of free indirect speech are used for other purposes. Consider (31)

-(32):

(31) a. A: Are you a student? b. B: Am I a student?

(32) a. A: Leave me alone. b. B: Leave you alone!

In (31), B repeats A's utterance, using free indirect speech. (Notice the sentence form and shifted person.) However, (31B) is not a report: B is asking a question about what he has heard. In the same way, in (32), B repeats A's utterance, and perhaps expresses a sarcastic attitude to what he has heard.

Direct and indirect speech and thought can also be used for other purposes

than reporting. Consider (33):

(33) a. Peter: I'm the best cook in England.

b. Mary: You say you're the best cook in England. c. Mary: You think you're the best cook in England.

(31)

know how they are recognised and how they may be appropriate or relevant. Any adequate theory of quotation has to provide some account of this.

As I noted in the introduction to this chapter, a quotation does not necessarily repeat the original verbatim. Consider (34):

(34) A: What did he say?

B: a. Well, he said there's no way to help me. b. He said that there was no way.

c. He said, 'Sorry, there is no way for me to help you with this matter.' d. He said 'There's no way.'

e. There's no way.

All the utterances in (B) can be used to quote the same original utterance. It follows that we do not have to reproduce it verbatim. This raises a number of questions which will be dealt with in more detail below.

Notice, first, that the degrees of looseness permitted or expected in quotations differ depending on the circumstances. For example, in academic writing, anything within quote marks is supposed to reproduce the exact words of the original writer (except where the quotation is a translation). In court, witnesses are expected to report what they heard as precisely as possible, whether directly quoting or not. By contrast, the exact words may not be necessary in some contexts. For example, we do not expect reported speech to be identical with the original when talking about the weather. Consider (35):

(35)

a. The weather forecast says that tomorrow will be better than today. b. The weather forecast said that today would be better than it is.

In (35a), it is not crucial whether the weather forecast said exactly 'Tomorrow will be better than today' or something like 'Tomorrow we will have more sunny weather' or 'Temperature will rise tomorrow'. A classic example where the indirect

(32)

CI apter I

the weather forecaster made the contradictory statements 'Tomorrow will be better than it is'. As in other cases we have seen, the reporter is paraphrasing in his own words, and perhaps elaborating as the original, and would expect to be understood as such.

Another case where there is no need for verbatim reproduction is when we quote someones utterance for his accent or attitude. Take an example:

(36) He must be from Scotland. He said 'Tell me where the park is'.

In (36), if the speaker uses a Scottish accent in the direct quotation, he would be justified in considerably simplifying the content of the original, which may add little to

what he wants to get across.

The assumption of verbatim reproduction is also undermined by reported speech in translations from other languages. Translation obviously does not preserve the exact words of the original; moreover, word-for-word translations are not always the most appropriate, and even semantic or logical identity may not be achievable. This is obvious to anyone faced with translated speech, whether direct or indirect. We merely expect the translation to be faithful enough.2

In this section, I have tried to show that reported speech and thought - like mention - may be used to draw attention to either the form or the content of the original. Its presence may be overtly marked or left to be inferred. I have also argued that quotations need not be verbatim reproductions of the original. This point will be discussed further below.

(33)

Chanter I

1.2.3 Mixed quotation

Mixed quotations are cases in which an expression is simultaneously used and mentioned (see Clark and Gerrig 1990; Cappelen and Lepore 1997a; Wilson 1997). Consider (37):

(37) The teacher used 'the rod of love' to make us learn better.

In (37), the speaker simultaneously asserts that the teacher used the rod and alludes to the teacher's (or some other person's) actual words. Normally, when an expression is directly quoted, its semantic content is considered immaterial to the interpretation process: the quoted expression may be in another language, or a nonsense form, without affecting the acceptability of the utterance. In mixed quotation, though, the semantic content of the quoted expression makes an essential contribution to the interpretation process. An adequate account of quotation should explain how this type of quotation is recognised, and how it contributes to the overall interpretive process.

Mixed quotation is also used in reported speech. Consider (38):

(38) The teacher said that he would use 'the rod of love' to make us learn better.

In (38), the speaker is indirectly quoting the teacher's utterance, and at the same time, directly quoting some of the teacher's words. The analysis of (38) is not so simple as that of (37). For example, the teacher himself may have quoted the words, as in (39a); he may merely have used them, as in (39b):

(39) a. The teacher said, 'I will use "the rod of love" to make you learn better'. b. The teacher said, 'I will use the rod of love to make you learn better'.

(34)

Chap ter 1

Some of the oblique expressions we looked at in the last section might be analysed as cases of mixed quotation. Consider (40):

(40) a. John said that the evening star is an inferior planet. b. John said that Venus is an inferior planet.

The de dicto reading (on which John used the words 'the evening star' in (40a) and

'Venus' in (40b)) might be seen as a case of mixed quotation. Thus, in (40), if the reporter thinks that John was talking about 'the evening star' without knowing that it is also called 'Venus', and this is critically important, she might mark 'the evening star' as directly quoted from John's utterance, as in 'John said that "the evening star" is an inferior planet'. If it is not so important, she may leave it to be inferred. The case of mixed quotation may therefore have more philosophical and semantic importance than has hitherto been realised.

There are some restrictions on the use of mixed quotation. Consider (41), repeated from (2) above:

(41) a. Socrates said, 'Know yourself!'

b. Socrates said that we should know ourselves.

(4 Ia) is a case of direct quotation and (41b) of indirect quotation. With examples of this type, mixed quotation, as in (42), is not normally considered acceptable (the same indices indicate the same referent):

(42) ?Socrates said that we, should know 'yourself1'.

Here, the directly quoted part 'yourself' cannot be coreferential with 'we'.

(35)

Chapter 1

(43) a. If Danny 1 's colleagues unanimously speak highly of him 1 , without showing a tinge of jealousy at his, success, 'maybe it's because, 1 1 'm harmless...' b. . . .she 1 herself cautions that the observations 'are based on examples that

readily occurred to me 1 , and are therefore inevitably subjective'.

I will try to explain this in chapter 2, section 2.4.3.

Another similar restriction is that there should be no clashes in agreement between the quoted and non-quoted parts of the sentence. For lack of agreement in number (and maybe in personal pronoun), (44) is not acceptable:

(44) ?He 'love my country'.

Finally, mixed quotations should be distinguished from both reported speech and thought on the one hand, and mention on the other. Consider another example, from Wilson (1997: 6):

(45) The outcome is 'a consummation devoutly to be wished'.

(36)

Chapter /

In this section, I have looked at various standard types of quotation: mention, reported speech and thought, and mixed quotation. The originals of these quotations may be mental representations (e.g. thoughts), public representations (e.g. utterances) or abstract representations (e.g. sentences and propositions); they may be linguistic or non-linguistic, and the properties singled out by the quotation may be either formal or logical. I have also tried to show that not all quotations are verbatim reproductions of the original, and that the degree of resemblance expected depends on the circumstances. Sometimes the presence of a quotation is overtly marked (e.g. by quote marks or a verb of saying), but at other times, it is left to the hearer to infer. This raises the question of how the hearer can recognise the presence of a quotation, the source and type of the original, and the degree and type of resemblance intended. In the following sections, I will survey some traditional theories of quotation, to see what light they can shed on these questions.

1.3 Traditional Analyses of Quotation

In this section, I will survey some traditional accounts of quotation, including philosophical analyses of the use-mention distinction, and analyses of reported speech and thought in traditional grammar, semantics and literary analysis. My main aim is to show that, whatever their intrinsic merits, none of these approaches offers an adequate account of how the hearer resolves the many indeterminacies left open when a communicator uses a quotation, and that a full account of how quotations are understood must make extensive appeal to pragmatics.

1.3.1 Definitions of

mention

vs.

quotation

In the literature, three different positions have been taken on the relation between

(37)

Chapter 1

'The connection between quotation on the one hand and the use-mention distinction on the other hand is obvious, for an expression that would be used if one of its tokens appeared in a normal context is mentioned if one of its tokens appears in quotation marks (or some similar contrivance for quotation).' (Davidson 1984: 79)

That is, what is quoted (placed inside quote marks) is mentioned. On this definition, which is used by Davidson, mention and direct quotation contrast with indirect quotation. (For discussion, see section 1 .3.2.2.)

There is a second, narrower, definition, on which mention is a subtype of quotation in which the original of the quoted material is an abstract linguistic expression. Other types of quotation involve attribution of an utterance or a thought to a source, as in reported speech and thought (Cappelen and Lepore 1997a). Another similar position is that only attributive quotation is called quotation, and non-attributive quotation is called mention. On these definitions, mention contrasts with reported speech and thought. (For discussion, see section 1.4. 1.)

Finally, a third possible position is that mention and quotation are different: quotation is the use of quote marks or other devices such as finger dancing, which overtly indicates mention, but also performs other functions. Since not all mentions are explicitly indicated, it follows that not all quotation is mention, and not all mention is quotation. On this account, quotation and mention may, but need not, overlap (Saka 1998). (For discussion, see section 1.4.2.)

(38)

two varieties of metarepresentation has led to a neglect of many further data which could shed useful light on the analysis of quotation.

1.3.2 Theories

of

the use-mention distinction

1.3.2.1

Quine (1965) and Garver (1965)

Quine is one of the philosophers responsible for drawing attention to the use-mention distinction. Consider his often-cited examples (Quine 1965: 23):

(46) a. 'Boston' is disyllabic. b. 'Boston' has six letters. c. 'Boston' is a noun.

Quine treats (46a) as a statement which ascribes to 'Boston' the phonetic property of being disyllabic, (46b) as ascribing to 'Boston' the property of having six letters, and (46c) as ascribing to 'Boston' the grammatical property of being a noun. According to Quine, each occurrence of 'Boston' in (46) is a name of the word 'Boston', just as 'John' is a name of a person.

Despite its important contribution to later accounts of the use-mention distinction, Quine's account has been widely criticised. For example, Garver (1965) claims that, while useful from a practical point of view, the use-mention distinction is neither exclusive nor exhaustive. In his view, to describe the use-mention distinction in Quine's terms is to suggest that there must be instances of pure mention, where the relation between name and referent is arbitrary. Consider (47) - (48):

(47) John is my friend.

(39)

C

In (47), 'John' is the name of a person, but the relation between name and referent is not transparent: we cannot work out from the name who the referent is. Analogously, Garver argues, if 'cat' in (48) is an instance of pure mention, it must function in the same way, on Quine's account. However, it does not: we use the information contained within the quotation marks in determining the referent of 'cat' in (48): we cannot recognise it as denoting the word 'cat' unless we recognise 'cat' to be a word. Garver concludes that it is very doubtful that there is such a thing as pure mention, because the relationship between the alleged name and the thing it denotes is not arbitrary.

Garver, following Quine, assumes that to mention a linguistic expression is to name it, and therefore claims that there is no such thing as pure mention. Later analysts (e.g. Davidson 1979) have drawn the opposite conclusion: since the relation between mentioned expression and referent is not arbitrary, they conclude that mention is not naming. So they concede Garver's objection to proper-name theory, but draw a different conclusion.

Garver also argues that the use-mention distinction is not exhaustive. He tries to show that there is a third mode of occurrence, involving reference not to the form (as in standard mention), but to its meaning. Consider (49):

(49) The meaning of 'courage' is - steadfastness in the face of danger.

According to Garver, in (49) the expression 'steadfastness in the face of danger' is neither used nor mentioned. It is not used because (49) does not mean that the meaning of 'courage' has the virtue of steadfastness in the face of danger. It is not mentioned because (49) does not mean that the meaning of 'courage' is a certain linguistic expression, 'steadfastness in the face of danger'. Garver distinguishes this third mode of occurrence using double slanting lines, as in (50):

(40)

Chapter 1

In (50), '//'steadfastness in the face of danger'//' means 'the meaning of "steadfastness in the face of danger". He concludes:

'The moral to be drawn is that the distinction, considered generally, is neither exclusive nor exhaustive, and that there are varieties of mention just as there are varieties of use. These considerations do not, of course, call for abandoning the use-mention distinction, but rather for paying more attention to varieties of use and mention, so that use of the distinction can be more effective and its misuse eschewed.' (Garver 1965: 238)

Garver, like Quine, assumes that mention, as the name of a linguistic expression, denotes only linguistic form. He therefore needs another mode, denoting meaning. Later analysts (e.g. Sperber and Wilson 1981) have argued that what is mentioned may be either an abstract linguistic expression (e.g. a word, a sentence) or an abstract logical representation (e.g. a concept, a proposition), and I will follow them on this. Later analysts have also taken seriously some of the problems with the view that mention is a type of naming. A particular problem for the name theory of mention, and more generally for the name theory of quotation, is raised by the fact that quotation does not always involve verbatim reproduction. Some alternative to the name theory must therefore be found.

1.3.2.2 Davidson (1979)

(41)

Chapter 1

Davidson starts by discussing the claims of Tarski (1956: 159) and Quine (1961: 140) that quotations are unstructured proper names of the quoted expressions

(the proper-name theory). (We saw in the last section that Quine analyses a mention

as a name of the mentioned word, where the relation between a name and its referent is arbitrary.) Davidson makes a similar objection to Garver's:

'On this view [the proper-name theory], there is no relation, beyond an accident of spelling, between an expression and the quotation-mark name of that expression. ... Nothing left, either, of the intuitively attractive notion that a quotation somehow pictures what it is about.' (Davidson 1984: 83)

That is, Davidson denies that there is an arbitrary relation between a quotation and the original. He also notes that on the proper-name account, quotational sentences do not have truth conditions, because what is contained within quotation marks, being name-like, has no significant semantic structure. So he rejects the proper-name analysis of quotation.3

According to Davidson, what is contained within quotation marks is not part of the semantic structure of the whole clause in which it occurs. This is the starting point of his demonstrative theory. The main claim of the demonstrative theory is that quotation marks are used to refer to a shape which lies outside the semantic structure of the containing sentence. The quotation marks should be read as 'the expression a token of which is here', or 'the expression with the shape here pictured'. What occurs inside the quotation marks is the token of the expression or shape in question.

Thus, consider (51):

(51) a. 'Alice swooned' is a sentence

b. Alice swooned. The expression of which this is a token is a sentence.

(42)

According to Davidson, (5la) should be understood as (51b), where the token of 'this' is supplemented by the speaker pointing to the token of 'Alice swooned'.

Davidson sees two advantages in the demonstrative theory: one is that it 'assigns a structure to sentences containing quotations': that is, in

(5lb), the second

sentence can have truth conditions because the quotation is now removed from it. The other is that it explains 'the picturing feature of quotation': that is, the idea that a quotation is like a picture of the original, in that what is inside the quote marks is presented as having the same shape as the original.

Davidson also sees it as an advantage of the demonstrative theoiy that it deals with mixed cases of use and mention, as in (52), from Davidson (1984: 92):

(52) Quine says that quotation '... has a certain anomalous feature'.

According to Davidson, (52) is a case where the quoting sentence happens to contain a token with the shape needed for the purposes of quotation: 'Such tokens then do double duty, once as meaningful cogs in the machine of the sentence, once as semantically neutral objects with a useful form'. (Davidson 1984: 92) He analyses (52) as communicating something like (53):

(53) Quine says, using words of which these are a token, that quotation has a certain anomalous feature.

(43)

Chapter /

supplemented by an inferential element of interpretation. By contrast, overt quotation

marks do uniquely determine the scope of quotation. Thus, on Davidson's account, the distinction between overt quotation and inferred quotation becomes blurred.

Another problem is that Davidson's account of mixed quotation, as it stands, can be applied only to utterances of the type in (53), where the source of the quotation is overtly marked. It is not clear how it would apply to a case like (54), repeated from (37) above:

(54)

The teacher used 'the rod of love' to make us learn better.

In (54), the quote marks can be seen as pointing at 'the rod of love', but there is no overt indication of whether the teacher used these words. As a result, no paraphrase of the form in (53) is available for (54), and it is not obvious how he would deal with

these examples.

A further objection to the demonstrative theory hinges on Davidson's claim that quote marks are demonstrative. We have seen in section 1.2 that quotations are possible without the presence of overt quote marks. In the demonstrative theory, the role of pointing, which is essential on Davidson's account, is performed by the quote marks, and without them, it is not clear what does the role of pointing. Davidson might argue that, in any case, quotation marks need to be supplemented by a gesture from the speaker, and that in the absence of quotation marks, the speaker's gestures must bear the whole burden. But this again underlines the need for an adequate pragmatic account of how semantic indeterminacies are resolved.

Finally, Davidson (1979) does not discuss direct quotation, but I assume that, given his treatment of mixed quotation, he would be prepared to extend his analysis to this. If so, there is a further problem, in that this analysis can only deal with the case of verbatim reproduction. Because he analyses what is inside the quotation marks as picturing the shape of the original, he appears to exclude looser degrees of resemblances, e.g. translations of direct speech. Davidson's account of indirect quotation will be discussed in section 1.3.3.2.

(44)

er

In this section, I have surveyed two traditional accounts of the use-mention distinction, and raised some questions about them. Some of these questions have been tackled in more recent accounts, which in turn raise further questions. These will be discussed in section 1.4.

1.3.3 Theories

of

reported speech and thought

By definition, reported speech is a case in which a previous utterance is quoted in order to report it. Note that not all repeated utterances are quotations. Consider (55):

(55) Peter: I was with Mary last night...

I

was with Mary last night.

Maiy: Peter was with me last night.

(45)

1.3.3.1 Accounts of reported speech in traditional grammar

In traditional grammar, reported speech is divided into two subtypes: direct speech and indirect speech (see Onions 1905; Jespersen 1924, 1954; Kruisinga 1925; Curme 1931; Quirk 1972; etc.). 4 Here, I will look in more detail at Kruisinga (1925) and Jespersen (1954) as representative examples.

Kruisinga (1925: Part II, 3, 194 - 197) uses the terms direct style and indirect

style, defining the former as a case where the original is repeated without change and

the latter as a case where the thought expressed by the original is repeated in a subordinated clause. Notice that the examples in (55) above apparently qualify as reported speech by this definition.

Kruisinga deals with free indirect speech as semi-indirect style, defining it as a

case of indirect speech without an introductory clause. He suggests that the tense of the verb or the presence of certain auxiliaries might act as diagnostics, as in (56) -(57) (p. 195):

(56) . . .Vibart was sure that Jasmine was misjudging her. No one, could have been

more auxious [sic. anxious] to help him find Jasmine...

[Mackenzie, Rich Relative ch. 6]

(57) And if Grandmamma should die? Why then in this great house she, Mary Flower, should be all alone.

[id. Seven Ages of Woman ch. 3.]

According to Kruisinga, the past tense of 'was sure' and 'could have been' in (56), and the auxiliary verb 'should' in (57), show that the utterances are indirect speech.

Finally, Kruisinga analyses echo questions as examples of free indirect speech.

Let us look at his examples (p. 298):

(46)

Chapter 1

(58) 'I like you!' returned Aunt Bel, nodding at him. 'Where do you come from? A young man who'll let himself go for small coin's a jewel worth knowing'.

'Where do I come from?' drawled Laxley.

[Meredith, Harrington ch. 16]

(59) 'I have found you a gown after all.' '\Vhere is it?'

'Where is it?' her uncle repeated. 'Why, waiting upstairs in your bedroom, of

course, for you to put it on

[Mackenzie, Rich Relatives ch. 12, p. 298]

In (58) - (59), the second question in each example is an echo question, which Kruisinga includes in the category of free indirect speech. By contrast, Banfield (1982) does not. This issue will be discussed in more detail in section 1.3.3.3.

Jespersen (1954) also divides reported speech into two subtypes: direct and indirect. According to him, direct speech is used when the reporter 'gives, or purports to give, the exact words of the speaker (or writer)', and indirect speech is used when she 'adapts the words according to the circumstances in which they are

now quoted' (p. 290). He treats free indirect speech as a subtype of indirect speech; he calls it represented speech, as opposed to dependent speech, i.e. ordinary indirect speech with an introductory clause.

Jespersen draws attention to several differences between direct and indirect speech. One is tense: direct speech maintains the original tense of a verb, but indirect speech backshifts it when the subordinating sentence is in the past. Another is person: direct speech preserves it, while indirect speech shifts it to the speaker's viewpoint. I have illustrated these aspects in more detail in section 1.2.2. A third difference is that direct speech does not change the form of the original sentence, while indirect speech does in cases where the original is a non-declarative. Consider (60) - (61):

(60) a. He said, 'Where do you come from?' b. He asked me where I came from.

(47)

(61) a. He said, 'Leave at once.' b. He told me to leave at once. c. He told me I was to leave at once.

In (60), the indirect speech in (b) has a different word order from that of the direct quotation in (a): it has the word order of declarative, although the wh-word is fronted as is typical for an interrogative. The verb of saying is also changed, from 'say' to 'ask'. In (61), the imperative form in (a) becomes part of the main clause, as in (b), or an embedded declarative form, as in (c).

As Jespersen points out, in free indirect speech, only tense and person are shifted, so the sentence forms of questions and exclamations remain unchanged. The examples in (62) - (63) are from Jespersen (p. 298):

(62) How could he bear ... ? (from 'How can I bear...?')

(63) What a nuisance it was ...! (from 'What a nuisance it is...!')

In (62), the original question and the free indirect speech have the same sentence form, but different person and tense. The same goes on in (63).

According to Jespersen, imperative forms are not often used as free indirect speech. Consider (61) again: the imperative form in (61a) can be indirectly reported either by (6lb) or by (61c). Jespersen points out that the embedded clause in (61c) is generally used in free indirect speech. However, he also notes that the imperative form may occasionally be retained. Consider (64) - (65) ((64) from Jespersen (1954: 299)), and (65) from Jespersen (1924: 160)):

(48)

Chapter 1

(65) she was growing a trifle impatient: [and thought:] if he wanted to sulk, let him, she didn't care.

In (64), both the imperatives with 'look' are used as free indirect speech, as can be seen from the backshifted tense 'was'. By the same token, in (65), 'let him' is a case of free indirect speech, as can be seen from the backshifted tense 'wanted' and 'didn't'.

In traditional grammar, most cases of free indirect speech are simply indirect speech of which the introductory clause is omitted. Especially when indirect speech is repeated, the same introductory clause tends to be omitted from the second occurrence (Jespersen 1924: 155; Curme 1931: 420). On this approach, free indirect speech is not regarded as a type of speech designed and specialised for a specific literary style. This contrasts with the view of Banfield (1982), which I will turn to in section 1.3.3.3.

In traditional grammar, direct speech is described as reporting the original sentence verbatim. This is also held to be true of indirect speech, with the exception of shifts in person, tense, or deictics; and these shifts must not lead to an alteration in content. However, as we have seen in section 1.2, quotations are not necessanly verbatim reproductions. Traditional theories of reported speech are too restrictive to allow for varying degrees of resemblance between quotation and original. They also give no adequate definition of reported speech. As we have seen above, not all repetitions of another utterance are reported speech; the speaker's intentions need to be taken into account. All this calls for an adequate pragmatic theory.

1.3.3.2 Semantic accounts of indirect speech

(49)

the scope of a direct quotation is not normally truth-preserving: 'He said, "I saw a bachelor" does not entail 'He said, "I saw an unmarried man". By contrast, in indirect quotation, we can draw some valid inferences, e.g. 'He said he saw a bachelor' would generally be taken to imply 'He said he saw an unmarried man.' In this section, I will explore some semantic accounts of indirect quotation, 5 the basic aims of which, as noted above, are to get the truth conditions of indirect quotations correct and to capture their inferential relations to other sentences.

The truth conditions of the embedded clause of indirect speech do not affect and are not affected by the truth conditions of the main clause. It follows that the truth conditions of the embedded clause should be treated as semantically independent of those in the main clause. The embedded clause has been variously treated as an independent proposition (e.g. Church 1950), an independent sentence (e.g. Carnap 1947, Quine 1956, 1960), an independent inscription (e.g. Scheffler

1954), or an independent utterance (e.g. Davidson 1968). I will comment briefly on

these different approaches, and then discuss Davidson's account in more detail, since it is one of the most influential accounts, and also one of the closest to the metarepresentational approach, which I will take in chapter 2.

Some accounts of indirect quotation treat them as propositions which are the objects of propositional-attitude verbs like 'believe', 'know', 'say', etc. Consider (66):

(66) a. John said that Peter loved Mary.

b. John expressed the proposition that Peter loved Mary.

On these accounts, (66a) is analysed as roughly equivalent to (66b). This analysis was rejected by Carnap (1947), and many later philosophers, not only because propositions were held to be philosophically dubious, but also on empirical grounds.

See, for example, Carnap 1947; Church 1950; Quine 1956, 1960: chapter 6; Scheffler 1954; Davidson 1968, 1975, 1976; Smith 1976; Reddam 1981; Baldwin 1982; Burge 1986; Schiffer 1987; Soames 1987, 1989; Bertolet 1990; Richard 1990; Seymour 1994.

49 BIBL

The semantics and pragmatics of misrepresentation in English - a relevance-theoretic approach

The Semantics and Pragmatics of Metarepresentation in English:

A Relevance-Theoretic Approach

Eun-Ju Noh

Thesis submitted in partial fulfillment

of the requirements for the degree of Ph.D.

University College London

1998

For my parents

Abstract

1986/1995). The main aim is to deepen the relevance-theoretic analysis of

Acknowledgements

Table of Contents

Introduction

Metarepresentation in Communication

1986/1995), metarepresentation is

Chapter 1

Analyses of Quotation

'Know

1.2.1 Pure quotation (Mention)

Kellogg's

Kellogg's

(5) MATCH 3 TO WIN OR MATCH 2 AND 'DOUBLE' TO DOUBLE YOUR WIN

'Kellogg's' and 'DOUBLE' or

(6)

'Kick the bucket' has three words.

(35)

1.3.1 Definitions of

vs.

1.3.2 Theories

the use-mention distinction

1.3.2.1

(5lb), the second

(54)

1.3.3 Theories

reported speech and thought

I

was with Mary last night.

Maiy: Peter was with me last night.

1954), or an independent utterance (e.g. Davidson 1968). I will comment briefly on